The situation
A distribution company running Salesforce for six years across roughly 140 users. Nobody currently on staff had been there when the org was built. Two admins had come and gone. A simple request — add a field to the Opportunity page and report on it — had been quoted internally at three weeks, and nobody could explain why.
The symptom leadership actually noticed was different: the pipeline report and the finance report disagreed by a material amount every month, and both teams had stopped trusting either one.
What we found
The assessment ran read-only for two weeks. The inventory turned up:
- Four separate automations firing on Opportunity save — two workflow rules, a Process Builder process and a record-triggered flow — two of which wrote to the same currency field in an order that varied by trigger context
- 317 custom fields across the standard objects, of which 106 held no data in any record created in the previous 24 months
- Nine validation rules that no longer matched the current sales process, three of which reps were routinely working around by entering placeholder values
- An integration to the ERP that had been silently failing on a subset of records since a field rename the previous spring
The report disagreement traced directly to the competing automations: depending on how a record was updated, one of two different values won.
What we did
Remediation ran in three phases over about ten weeks, each one scoped and priced separately so leadership could stop after any of them.
Phase one consolidated the Opportunity automation into a single record-triggered flow with explicit ordering, then backfilled the affected records so historical reporting matched going forward. Phase two retired the dead fields and validation rules, working through a review list with each department so nothing got deleted that someone quietly depended on. Phase three rebuilt the ERP integration with a written field map and failure alerting, so the next silent break wouldn’t stay silent.
Outcome
The two reports reconciled and stayed reconciled. Field count dropped by roughly a third. More usefully, the org came out of it with a document describing what runs, what it does and who asked for it — which meant the ops manager could scope routine requests herself instead of guessing.
They moved onto ongoing admin support afterward, largely to keep the org from drifting back.